mpra paper 15021
DESCRIPTION
MPRA Paper 15021TRANSCRIPT
-
MPRAMunich Personal RePEc Archive
The Impact of Schooling Reform onReturns to Education in Malaysia
Ramlee Ismail
University Pendidikan Sultan Idris
29. January 2007
Online at http://mpra.ub.uni-muenchen.de/15021/MPRA Paper No. 15021, posted 9. May 2009 07:34 UTC
-
The Impact of Schooling Reform on Returns to Education in Malaysia
Ramlee Ismail
Universiti Pendidikan Sultan Idris
Tanjung Malim, Perak, Malaysia
Ramlee Ismail is a senior lecturer at Universiti Pendidikan Sultan Idris, 35900 Perak,
Malaysia. Email: [email protected]
-
ABSTRACT
The objective of this paper is to examine the impact of education reforms on earnings. One of
the significant changes in the Malaysian education system was the schooling reform of 1970
that changed the medium of instruction from the English language to the Malaysian national
language. Using data from the Household Income Surveys of 2002 and 2004, this paper
updates the private rate of return to education. Applying a homogenous return model, using an
ordinary least square (OLS) regression indicates that the private rate of returns to education is
close to the world average. Using the Instrumental Variable approach, however, the impact of
the schooling reforms indicates that the private rate of return to education is higher than the
average.
JEL Classification: C13, I21, J24
Keywords: human capital, instrumental variables, rate of return to education
-
1
1. INTRODUCTION
Human capital development is a prerequisite to the preparation of Malaysia to become a
knowledge-based economy and for sustaining economic growth. The capability and capacity
in the management of new knowledge and technologies will be determined by the quality of
its human capital. With globalization, this country will be facing more competition in trade
and investment. Therefore, the workforce will have to be equipped with a strong base in
education and training. It is also important to acquire a range of generic skills, such as
communications and analytical skills. In addition, a successful entry into the information age
will enable the economy to take advantage of the opportunities arising from the information
and technological revolution. However, the future will depend on a dynamic and responsive
education and training system to respond to global change. Education will be crucial in the
creation of a knowledgeable manpower to support new industries and economic activities, and
to develop an information-rich society. Priority should, therefore, be given to increasing
accessibility to quality education and training as well as to strengthening the human capital
base to support the development of a knowledge-based economy during the National Vision
Policy (NVP), 2001-2010. The education system in Malaysia has changed gradually to meet
the nations needs and aspirations since independence in 1957. One of the significant changes
was the school reform of 1970 when the English language was substituted by the Malaysian
national language as a medium of instruction in the government schools.
The objective of this paper is to update estimation of the private rate of return to
education in Malaysia, by estimating the average return for an additional year of schooling.
Furthermore, this estimation will provide new evidence of returns by using the latest data sets.
However, the main objective is to clarify the difference in the returns to different individuals
due to the impact of the schooling reform. In addition of the Ordinary Least Squares (OLS)
method by earlier studies, he alternative method, i.e. the Instrumental Variable (IV) is also
-
2
used, providing a new estimate of the returns for those who were affected by the educational
reforms. This method is increasingly important in the literature because it also reduces the
potential bias. Furthermore, it has never been applied to the Malaysian data.
The paper is organized as follows. The second section discusses previous studies of
returns to education in Malaysia, followed by an account of the Malaysian education system
and its reform. The fifth section is revealed the method of the study. The results are explained
in the next section and, finally the conclusion.
2. REVIEW OF LITERATURE
The studies of returns to education in Malaysia can be divided into two categories: those
which used a variety of data collected by official bodies or field surveys by researchers, and
those which utilized official data from the government such as the Malaysian Family Life
Survey 1 and 2 (MFLS1 and MFLS2) and the Household Income Survey.
In the first category, Hoerr (1977) conducted the first cost-benefit analysis of
education in Malaysia in 1973, using the Malaysian Socio-Economic Sample Survey of
Households, 1967-68. However, his study covered only a relatively small sample of 800.
Nevertheless, it was an important benchmark in investigating the returns to education in
Malaysia. His findings showed that the cumulative private rate of return to education was
higher for upper secondary education at 17.6 percent compared to primary or higher
education, which were 12.9 and 16.0 percent respectively. Mazumdar (1981) used the 1970
Post Enumeration Survey (PES) and World Bank Migration and Employment Survey (MES)
in 1975, which covered a small sample of male wage-earners and self-employed workers
using information from three urban areas. Lee (1980) used non-random samples of 1,179
from the private sector and 792 samples from the public sector employees in Klang Valley.
-
3
These studies concluded that the earnings variation in human capital theory is largely
explained by education. Chapman and Harding (1985) found that the average return to
education was 9.37 percent. Unfortunately, these figures did not represent the Malaysian
population as a whole but might be true for the return for their samples, which covered less
than a thousand respondents.
Other studies estimated the returns to education using MFLS 1 and 2. This survey was
conducted for the purpose of gathering data on economic and biological aspects of fertility
rates and other related variables. It was conducted by the government during 1976-79
(MFLS1) in Peninsular Malaysia. The sample consisted of 1,262 households in which at least
one married woman was aged less than 50 years at the time of survey. It also included the
earnings and occupational histories of the women, and the data for their husbands. Blau
(1986), Gallup (1997) and Chung (2004) estimated the rate of return to education using these
data. However, the results of their studies were inconsistent, probably because their objectives
and methods were different. The average rate of return to an additional year of schooling
education reported by Gallup was 7.6 percent. On the other hand, Blau and Chung did not
report the overall return. Chung (2004) has estimated that the marginal returns to education
were 12 percent for lower secondary, 17 percent for upper secondary, 26 percent and 17
percent for pre-university and higher education respectively.
Chung (2003) estimated the rate of return to schooling in Malaysia using a larger data
set, the Malaysian Household Income Survey 1997. She found that the marginal gross return
was higher at the upper secondary to pre-university level where an individual has an annual
gross return of 22.9 percent. This result is consistent with the previous findings but contrasts
with the study carried out by Hoerr. However, due to the many differences in the sample and
estimation, a comparison between the earlier and later studies is difficult. For example, the
-
4
study conducted by Chung included more explanatory variables, such as number of wage
earners in the household, self employed, marital status and gender.
The estimated returns to education were different for two main reasons. Firstly, the data
sets in the studies are different. Secondly, the method or model specification was not the same
despite most of the studies having used OLS as a tool of analysis. The limited data and
resources, and to some extent the choice of schooling and earnings variables, also result in
different estimated returns to education. Moreover, some of the studies, such as Gallup (1997)
and Mazumdar (1981), emphasized income inequality rather than return to schooling. The
explanatory variables in the earnings equations are also quite different. Blau (1997), for
example, included occupational dummies which has an impact on the schooling coefficients.
Nevertheless, this is common when researchers have different sets of potential regressors in a
data set. The dissimilarity of methodological aspects and data availability, economic and
educational change will affect the outcomes. On top of that, the results may be biased due to
measurement error, omitted variables or the absence of information about how ability affects
schooling choice. Nevertheless, these studies that estimate the returns to education have made
a great contribution to the literature relevant to Malaysia.
Our study is ask to estimate the rate of return for a later period. The Malaysian
education system has been going through changes over several decades. Before proceeding to
the estimation, therefore, we describe the structure of the education system the next section,
followed by an account of the reforms.
-
5
3. THE MALAYSIAN EDUCATION SYSTEM
Currently, the Malaysian education system consists of pre-school, primary school, secondary
school and higher learning institutions. The main purpose of pre-school is to provide a basic
education for young children before they go on to formal education. The objectives of pre-
school education are to foster love for the country, instil moral values, develop character,
develop basic communication skills and respect for the national language, acquire the basics
of the English language, appreciate physical activities and, finally, develop critical thinking
skills through enquiry and the use of all the senses (Ministry of Education, 2001). Pre-school
education begins at the age of 5 or 6 at a government kindergarten, a non-government agency
or a private sector kindergarten.
Primary education starts at seven and continuous for six years. The structure of primary
education in Malaysia can be divided into two phases. The first phase is from Year One to
Year Three and the second phase is from Year Four to Year Six. During the first phase,
students will go through the curriculum to master the 3Rs; i.e. Reading, Writing and
Calculating (Arithmetic) to be used in daily life. In the second phase, i.e. from Year Four to
Year Six, mastery of the 3Rs is reinforced and emphasised by acquisition of general
knowledge, pre-vocational education, and the development of personality, attitude and social
values as well. Over the six years of primary education, students are assessed by continuous
school-based assessment until, at the end of Year Six, they experience the first National
Examination known as the Primary School Achievement Test (PSAT) to evaluate their
performance. All students are automatically promoted to secondary school after completion of
six years in primary school.
The normal duration of secondary schooling is five years but it is divided into two
levels. Level one refers to Forms 1, 2 and 3 (Lower Secondary) and level two refers to Forms
-
6
4 and 5 (Upper Secondary). Under the New Integrated Secondary School Curriculum,
secondary schools offer a comprehensive education programme with a wide range of subjects
from the arts and sciences to vocational and technological education with a practical basis.
Students in the government schools must sit two national examinations at the end of each
level; namely Lower Secondary Examination (LCE) at the end of level one, and Malaysian
Certificate of Education (MCE) after finishing level two. The Upper Secondary Education
offers choices to students to fulfil their needs, skills and interests in career development. All
Malaysian government schools use the same curriculum known as the Integrated Secondary
School Curriculum. Besides these schools, another choice is to enter Technical and
Vocational Schools which offer core and elective subjects in various technical and vocational
combinations. The purpose is to prepare students to pursue their study in technical and
engineering tertiary education, or to enable them to take up a career as technical and semi-
skilled workers. They have two years to prepare themselves for the third national
examination, which is the Malaysian Certificate of Education (MCE). Post-secondary
education offers school leavers or students the opportunity to continue their studies after
completing five years of secondary education. The options in post-secondary education are
not only in the academic field but also in various studies including matriculation, technical
and vocational, and short term courses. These courses are conducted by government and non-
government agencies, or in the private sector. Form 6 education is a continuation of the five
years of academic schooling that helps students to prepare themselves to qualify for
university. It takes two years to complete the post-secondary education either in the science or
the arts stream before the student can sit for the Higher School Certificate (HCE), conducted
by the Malaysian Examination Council. This education systems has worked as a result of a
series of reforms, notably in terms of language of instruction. Those reforms are explained in
the following section.
-
7
4. SCHOOLING REFORMS
The post-independence era was the starting point for the foundation, continuous change and
development of the Malaysian education system. The early years of independence were the
period of reconstruction intended to build the nation in the Malaysian mould. At that point of
time it was thought to be very important to integrate the multiracial society and to build up a
strong nation. The basis of that unity was to be laid by the school and education system. It
was an important objective of the education policy to bring together all races by gradually
making the Malaysian language the medium of instruction, as addressed in the Razak Report
of 1957. This report was reviewed by the Review Committee (known as Rahman Talib
Report, 1960), which suggested that the public accept the education policy proposed by a
previous report. The recommendations from both reports were important sources for the most
significant shift in Malaysian education that led to the implementation of the new Education
Act in 1961. The act also provided comprehensive and universal free education whereby all
students were granted automatic promotion up to Form 3 (Grade 9) in secondary schools
(Ministry of Education, 1980).
The first impact of the changes was the upgrading of the various types of primary
schools to national schools. Subsequently, gradual implementation of the Act has seen the
overall changes from the British education system to the Malaysian education system, with a
Malaysian outlook and orientation. The second impact of the legislation was the introduction
of the Malaysian language as the official medium of instruction in all government schools. It
was started in Primary 1 in 1970, and continued thereafter. At the end of 1978, all schools
were using this language as the medium of instruction and in the mid-1980s the universities
followed suit. This was a significant change in the Malaysian education system. The adoption
of the Malaysian language at all levels was considered necessary to ensure that the education
-
8
system became a tool for the integration agenda as addressed in New Economic Policy, 1971-
1990. It also aimed to promote nationhood and national identity starting from the grassroots
level (Neville, 1998). On top of that, the school reforms would give better opportunities to
people in rural settlements and to poor families in the enhancement of their level of schooling.
Furthermore, it was seen as the main tool to be used in the eradication of poverty, narrowing
and eventually closing the education gap between regions and races, as well as integrating the
education systems of the Sabah and Sarawak states with the national system (Okposin et al.,
2005).
Since the 1970s, the education system also reflected the changes in the needs of the
labour market in which there was great emphasis on science and technology. Technical and
vocational courses were also popular due to the higher demand for skilled and semi-skilled
labour. The curriculum also changed tremendously by adapting the syllabus to the changing
needs of the nation, especially the adapting of the curriculum to fulfil the development needs
of the country. The last decade of the twentieth century witnessed an extraordinary and
accelerating change in the Malaysian education system. Due to liberalisation, the
globalisation process and advances in information technologies, the Malaysian education
system has had to maintain a pace parallel to the international process. A balanced and
integrated approach has been taken to make sure that the nation is not left behind in terms of
technological development. The country should move at the same pace as other countries and
should also take hold of the emerging opportunities of new technologies, economic and social
progress, by re-structuring and re-focusing, as well as reforming, its education system towards
the markets needs, and to meet global competition. In order to do this, some changes had to
be made and, accordingly, several adjustments were carried out such as the Education Act
1961 being replaced by the Education Act 1996. Furthermore, some educational legislation
was enacted and amended to support the new aspiration to achieve a developed nation by
-
9
2020. The important legislation educational institutions are University and Universities
Colleges 1996, Private Higher Education Institution Act 1996, National Accreditation Board
1996, National Council on Higher Education 1996 and National Higher Education Fund
Board 1996.
In estimating the returns to education, we will consider a period when those reform
were all in place. We turn now to discussion of the method of estimation.
5. THE METHOD
The empirical analysis of this study uses a human capital earnings function to estimate the
private rate of return to education in Malaysia. Since the breakthrough by Mincer the earnings
function has been widely used to estimate the returns to education. The empirical model used
in this study starts from the Mincerian earnings function that is already known in the literature
as a benchmark and will use this to estimate the average private rate of returns to education in
Malaysia. The basic specification is:
iiiii ExpExpSW ++++=2
211ln (1)
where ln iW is log earnings, iS is years of schooling, iExp is the potential experience of
individual i , and i is a well-behaved error term. The last term of the equation, 2
iExp
represents experience squared to capture a concavity of the observed earnings profile. Due to
the absence of complete data on experience, Mincer (1974) proposed the alternative of
potential experience, i.e. the number of years individual A could have worked after
completing schooling. Assuming that he/she starts schooling at 7 years old and begins
working immediately after iS years of schooling, iExp is equal to AS7 (Age Years of
Schooling 7).
-
10
Applying simple Ordinary Least Squares (OLS) to the above equation, one can estimate
the coefficient 1 as the average of the private rate of return to schooling. The estimation of
the parameters 1 and 2 are generally positive and negative respectively. Mincer (1974)
claimed that weekly earnings were preferred as a dependent variable in the model. His
argument was that individuals with more education tend to work more and will receive higher
earnings compared to those with less education. However, in the literature on the human
capital earnings function, a variety of earning measurements have been used to estimate the
rate of return. For example, the alternatives of annual or monthly earnings have been used as
the dependant variable, depending on data availability. Consistently, the earnings variable in
equation (1) makes use of the logarithmic form because the distribution of log earnings is
very close to a normal distribution, especially log hourly wages (Card, 1999). In addition, it is
preferable to use the log transformation based on the success of the standard (semi-logarithm)
human capital earnings function (Willis, 1986). The method used here is preferable having
regard to the data available and the log transformation is convenient for interpretation in this
study. Therefore, this study uses monthly earnings as the dependent variable.
Despite the popularity of using OLS with the Mincerian earnings function, its use raises
a number of issues regarding the robustness of estimation. OLS regression of log earnings on
schooling will produce a bias in estimation on 1 because of the correlation between
iS and i . The sources of bias could emerge from three sources. Firstly, returns bias occurs
because of the correlation between marginal returns with the schooling choice of iS . It is not
very clear, but depends on the average returns among the sub-population of those with iS .
Schooling may be endogenous as a result of the individuals optimal schooling choice.
Consequently, OLS estimates will be biased upward. Secondly, ability bias is due to the
unobservable factor that is correlated with both schooling and wages, also leading to
estimation bias. Moreover, if ability is believed to be associated with both wages and
-
11
schooling (Ashenfelter at el., 1999), estimates of the return to schooling will tend to be biased
upwards (Griliches, 1977; Card, D 1999). However, most of the cases of omitted ability are
biased by not more than 5-15 percent (Schultz, 1988). Finally, a third source of potential bias
is associated with the measurement error. This bias, associated with schooling measurement,
age and experience, is misreported in the data. The simple way to deal with this problem is to
include the omitted variable in the equation. This means that ability becomes an explanatory
variable in the equation. Nevertheless, it must be taken into consideration that ability itself is
also influenced by schooling; hence, using the proxy, this variable will be biased downwards.
But recently most researchers have used Instrumental Variable (IV) estimation to avoid this
bias, although there is still no consensus about the correct approach.
The impact of schooling reform on the private rate of return to education in Malaysia
can be estimated using the IV approach. Using IV, participants can be permitted to self-select
into treatment and control, and one can subsequently tease out the exogenous impact using the
instrument. The IV operates by constructing another variable, which is not correlated with
earnings but is correlated with educational attainment. This should lead to a consistent
estimate of the rate of return. The general endogenous schooling model consists of the two
equations:
iiiii SXW ++='
ln (2)
where
iii ZS +=' (3)
In equation (2), iWln is determined by a vector of exogenous variables iX and years of
schooling iS . Meanwhile, the i s are interpreted as estimates of the private rate of return to
education. Estimation of the equation (2) by OLS will yield consistent estimates of i if the
-
12
iS is exogenous, so that there is no correlation between the two error terms. If this condition is
not satisfied, alternative estimation methods (i.e. IV approach) must be employed since OLS
will be biased. The model is a reduced form in which providing variable in vector iZ that is
not contained in iX (Pons & Gonzalo, 2001). That is a vector of exogenous variables which
influence schooling that can legitimately be omitted from the earnings equation. Then, replace
the schooling in equation (2) with the predicted or fitted value for schooling.
The estimation also known as two stage least squares (2SLS), which operates using
two steps. First, estimate the effect of the IV variable (school reform of 1970) on schooling
and, then estimate the effect of the instrumental variable on earnings. This is based on the
assumption that the school reform is correlated with earnings only because it influences
schooling, so the ratio of the effect of the instrument on earnings to its effects on schooling
will provide an estimate of the causal effect of school reform on earnings (Ashenfelter et al.,
1999). Many researchers apply IV estimation with different types of policy reforms to
estimate returns to schooling and compare the results with those derived using OLS. For
example, Harmon and Walker (1995) used the change in the school leaving-age (SLA) in UK,
which first occurred in 1946 from 14 to 15, and then from 15 to 16 in 1973.
The exogenous impact on the Malaysian education system was the introduction of the
Malaysian language as the official medium of instruction, and this is the instrument chosen in
this study. Under these circumstances, those students born after 1963 automatically used the
national language in the learning process. D70, is thus a dummy variable which is equal to 1
for individuals starting schooling in 1970 and thereafter, and otherwise is equal to 0. Given
the year of the reform, affected individuals (Zi = 1) are taken to be those who were born in
1963 and later. This exogenous variable affected the decision and opportunity to pursue
education at higher levels. In this context, IV estimates of the return to schooling using a
medium of instruction reform as the instrument, would be interpreted as the average return to
-
13
schooling for those who were affected by the policy reform. Borrowing the terminology from
the literature on treatment effects, Zi (exposure to different education system reform) is
independent of individual ability and the reduced form schooling residual (Heckman &
Vytlacil, 2000), with the assumptions that there is heterogeneity in the returns to schooling
and that the IV estimate is the Local Average Treatment Effect (LATE) (Imbens & Angrist,
1994; Blundell et al., 2000; Blundell et al., 2004).
6. DATA AND RESULTS
This study uses data from the Malaysian Household Income Survey (HIS) for the years 2002
and 2004. It is provided by Economic Planning Unit (EPU), Prime Ministers Department,
Malaysia. HIS2002 covered about 37,763 households in Malaysia. 11.42 percent or 4,313
observations from this survey were dropped from the estimation as not being in the labour
force. It also excludes people with no income at the time of survey. Those with extraordinary
earnings, i.e. more than MYR50,000 per month are also excluded. For HIS2002, only 5
observations earned an amount equal to or more than this. Students, pensioners, housewives
and unpaid workers were also not included. This group consists of 3,760 observations from
the whole population. The final sample of HIS2002 is 13,324 observations or approximately
35.29 percent of the total heads of households in the surveys. The HIS2004 included
information from 36,481 household heads. Initially, 22.19 percent of these observations were
dropped from the population because they were not in the labour force. Next, the pensioners,
students, home makers and unpaid workers were excluded. This left 13,492 from the
HIS2004, approximately 36.98 percent from the total of household heads in HIS2004. The
descriptive statistics are shown by Table 1.
-
14
Table 1: Descriptive Statistics
Years
Mean 2002 2004
Monthly Income (MYR)
Pooled 1974.34 2063.74
Urban 2290.73 2551.39
Rural 1486.88 1474.61
Employees 2069.26 2129.34
Self-employed 1750.00 1915.20
Schooling (years)
Pooled 9.03 9.04
Urban 9.85 10.03
Rural 7.78 7.84
Employees 7.78 7.84
Self-employed 7.29 7.30
Certificate
Pooled 2.19 2.24
Urban 2.45 2.56
Rural 1.78 7.84
Employees 2.43 2.49
Self-employed 1.62 1.68
Age (years)
Pooled 40.27 40.93
Urban 39.05 39.48
Rural 42.14 42.68
Employees 38.09 38.89
Self-employed 45.42 45.54
Experience (years)
Pooled 24.23 24.89
Urban 22.20 22.44
Rural 27.36 27.84
Employees 21.31 22.09
Self-employed 31.13 31.24
Overall Sample 13,324 13,492
The mean monthly income in 2002 was MYR1974.34 and increased to MYR2063.76
in 2004. As compared between strata, monthly income for those who had settled in the rural
areas decreased slightly from MYR1486.88 to MYR1474.61 during 2002 to 2004.
Meanwhile, the monthly income for urban areas in year 2002 and 2004 is MYR2290.73 and
MYR2551.39 respectively. These figures show that the income gap between strata has
widened. Meanwhile, earnings for those who were employed were 18.24 percent higher than
those who self-employed or employers in 2002. The earnings gap between these two groups,
however, declined in 2004. Employees received only 11.18 percent higher than self-employed
in 2004. In absolute figures, they earned about MYR1915.20 and MYR2129.34 (in current
-
15
price) in 2002 and 2004 respectively. The earnings gap between employees and the self-
employed decreased during this period.
The means for schooling, certificate obtained, age and experience increased slightly
during years 2002 to 2004. Age and experience have been increased by two years during this
period. On the other hand, comparisons between groups reveal a huge difference in the levels
of education, with persons in urban areas more likely to be better educated. For example, in
2004, the mean of schooling for urban areas was 10.03 years but for the rural areas only 7.84
years. The mean certificate obtained was different between these two groups by almost 1
point. The gap of the mean educational attainment between rural and urban samples is very
noticeable. The mean of schooling in 2004 for rural samples (7.84) doesnt reach the mean
figure of urban samples for year 1995 (9.16). In 2002, years of schooling for the self-
employed and employed was 7.29 years and 9.78 years respectively. It differed by 1.69 years.
The figures increased to 7.30 years for the self-employed and 9.81 years for paid workers
respectively in 2004. The mean difference between these two groups in 2004 decreased to
almost half a year. The mean certificate also shows the same trend during this time of period.
The educated workers are more likely to receive earnings in wages, and participate as an
employee in the labour market.
The mean age for the pooled sample in 2002 was 40.27, which increased to 40.93 in
2004. The mean of experience increased from 24.23 years in 2002 to 24.89 years in 2004. The
mean age for the urban sample was 39.05 years, and 42.14 years for the rural in 2002. In
2004, the mean age for the rural sample was 42.68 (increased by 1 year) and 39.48 for the
urban sample. The mean of experience was 22.20 years in 2002, and increased to 22.44 years
in 2004. In the meantime, the mean of experience of the rural sample was about 27 years in
2004. Obviously, the samples indicated that persons from the rural areas were older than the
urban by more than 2 years. However, the age difference within groups was obvious for the
-
16
self-employed and employees. For HIS2002, the mean age for employees was 7 years less
than the self-employed (45.42 years for self-employed and 38.09 for employee). The same
trend was found for HIS2004. In terms of experience, the mean for the self-employed in 2002
was 31.13 years but only 21.31 years for employees. These figures were raised to 31.24 and
22.09 in 2004 for the self-employed and employees, respectively. Interestingly, the samples
have shown an enormous difference between the means of age and experience between the
self-employed and employees, where employees were younger than the self-employed.
Moreover, the mean of experience for the self-employed was 10 years greater than for
employees not only for HIS2002, but it also for HIS2004.
Given this heterogeneity in the sample, we will use the IV approach to estimating rate
of return to education in Malaysia. But we will also generate OLS estimates for comparison
with other studies.
These data were subjected first to the same kind of OLS analysis used by various
studies for Malaysia. The return to education in the homogenous return model is constant
across individuals. The empirical results were derived from the estimation using equation 1 as
presented by Table 2. Column 2 and 4, reported the OLS estimates for year 2002 and 2004,
respectively. It estimated the Mincerian earnings equations where the natural log of monthly
earnings received by an individual is a function of years of schooling, potential experience
and its square, while the control variables used dummies for gender, marital status, household
heads activities and location (settlement type and zone of residential).
-
17
Table 2: The Private Rate of Returns to Education, 2002-2004
2002 2004
Variables OLS IV OLS IV
Schooling .1051**
(.0018)
.1174**
(.0049)
.1004**
(.0018)
.1109**
(.0043)
Exp .0381**
(.0016)
.0394**
(.0017)
.0292**
(.0016)
.0298**
(.0016)
Exp2 -.0005**
(.0000)
-.0004**
(.0000)
-.0003**
(.0000)
-.0002**
(.0000)
Female
-.1037**
(.0178)
-.1061**
(.0178)
-.0859**
(.0178)
-.0873**
(.0180)
Single .1140*
(.0165)
.1069**
(.0165)
.1513**
(.1646)
.1453**
(.0173)
Widow -.0794*
(.0311)
-.0757*
(.0312)
-.0074
(.0311)
-.0033**
(.0292)
Divorced -.0825**
(.0357)
-.0776*
(.0359)
-.0400
(.0357)
-.0389
(.0368)
Employee .0753**
(.0126)
.0697**
(.0129)
.0179
(.0126)
.0129**
(.0127)
Rural -.2392**
(.0099)
-.2270**
(.0109)
-.2994**
(.0099)
-.2885**
(.0108)
Central .1306**
(.0149)
.1212**
(.0153)
.0939**
(.0149)
.0871**
(.0159)
East -.3027**
(.0148)
-.3047**
(.0149)
-.2563**
(.0148)
-.2575**
(.0153)
North -.2018**
(.0139)
-.2055**
(.0324)
-.1886**
(.0134)
-.1915**
(.0144)
Sabah & Sarawak -.0895**
(.0154)
-.0801**
(.0359)
-.1551**
(.0154)
-.1465**
(.0157)
Constant 5.8371**
(.0319)
5.6885**
(.0648)
6.0672**
(.0319)
5.9405**
(.0579)
R-squared 0.3937 0.3913 0.3893 0.3875
F 618.39 409.47 570.26 407.77
Test Result
Partial R2 for excluded
variable instrument at first
stage
0.9311
(0.000)
0.9173
(0.000)
F-test
[p-value]
9644.73
[0.000]
7534.30
[0.000]
Endogeneity test-Wu
Hausman
F-test
[p-value]
7.5676
[0.0059]
7.3390
[0.0068]
Chi-sq 7.5714
(0.0059)
7.344
(0.0067)
Observations 13,324 13,324 13,492 13,492 Robust standard errors in parentheses.
** Significant at 1 % level. *Significant at 5 % level.
The average private rate of return for an additional year of schooling was 10.51 percent in
year 2002 and 10.04 percent in 2004. One additional year of experience increased earnings by
4 percent in 2002 and 3 percent in 2004. With the exception of the dummy for employee (in
-
18
2004), all parameters are significant at 0.05 levels or better in all years. Most of the
coefficients are significant at the 0.01 level. The results show the Malaysian data are
consistent with the basic human capital model. Regression on earnings function by controlling
gender, marital status, activity and area of residence give results that are in line with the basic
theory. Schooling and experience are positively correlated with earnings but experience
squared is negatively correlated.
The average return to education based on a homogenous return model (OLS) for
Malaysia is consistent with the average return for middle-income countries, which is 10.7
percent (Psacharopoulos & Patrinos, 2002) and slightly higher than the Asian average. The
private rate of return for Asia as a whole in 2004 was 9.9 percent (Psacharopoulos & Patrinos,
2004). Nevertheless, it is low compared to the Asian Tigers. For example, Singapore with an
average return of 13.4 percent in year 1974 (Psacharopoulos, 1994) and 13.1 percent in 1998
(Sakellariou, 2003); the Republic of Korea from 12 to 13.5 percent between 1974 and 1986
(Ryoo et al., 1993). But in Thailand, which is similar in terms of economic development, the
private return almost equals the return for Malaysia. For example, an average return in
Thailand (Hawley, 2004) was estimated at between 10.3 and 10.7 percent from 1985 to 1998.
Both Malaysia and Thailand enjoyed considerably higher returns compared to the rest of
Southeast Asia. In Vietnam, for example, average returns from education for an additional
year of schooling were 4.8 percent for the overall sample, and 3.4 and 6.8 percent for males
and females, respectively (Moock et al, 2003), whereas in Indonesia young people benefited
slightly more those than those in Vietnam from an additional year at 7.0 percent in 1995
(Duflo, 2001).
Now we consider the heterogeneous returns model, i.e. IV approach. This is the first
application of this approach to Malaysia. The first step of estimation is to examine the
relevance and validity of the instrument. The strong correlation between dummy D70 with
-
19
endogenous variable (schooling) and orthogonality to the error process needs to be confirmed.
Otherwise, the results will be biased and inconsistent. The degree of correlation to the
endogenous variable is tested by examining the fit of the first stage equation which included
the dummy D70 (Bound et al., 1995; Patrinos & Sakellariou, 2004). The results of tests using
a dummy year of changing the medium of instruction in schooling are statistically significant.
The F-test is equal to 9644.73 and p-value is 0.000 for HIS2002 and 7534.30 (p = 0.000) for
HIS2004. With regard to the quality of the D70s dummy, the F-test on excluded variables
and partial R2, is reported in the first row under Test Result at the bottom of the Table 1
(Column 3 and 5). Furthermore, the robust regression approach is used in case
heteroskedastic errors are present.
Results from both the 2002 and 2004 reduced-form equation of schooling have a
highly significant effect on length of schooling and no direct impact on earnings. In other
words, all equations are exactly identified. This is evidence that the D70 dummy can be used
as a valid instrument for schooling. In addition, any potential endogeneity in schooling was
also being checked. Using the well-known Durbin-Wu and Hausmans test, the hypothesis
that the OLS estimates differ is accepted at the significance level of 1 percent. All diagnostic
tests of relevancy and validity having been satisfied. D70 was therefore acceptable as the
instrument for IV. All the diagnostic test results are presented in the bottom rows (Test
Result) in Table 1. By obtaining the original controlling variables, dummies for gender,
marital status, activity and region (zone), the results suggest that IV estimates were somewhat
higher than those derived using OLS. Column 3 and 5 provide the rate of returns estimated
using IV, at 11.74 and 11.09 percent for 2002 and 2004, respectively.
The private rates of returns to education by IV estimation are approximately 11.70 and
10.46 percent higher than those resulting from the use of OLS. It is frequently found in the
literature that the standard error from IV estimation is higher than that from OLS (for
-
20
example, see Card, 1999 & 2001). These results are in line with Brunello & Miniaci (1999)
for Italy. They used data from male household heads drawn from The Bank of Italy Survey
(from 1993 to 1995). The important exogenous event in Italian education, which is Law 910
of December 1969, was used as the instrument. Their results suggested that the private rate of
return increased from 4.8 percent (OLS) to 5.6 percent (IV). It was higher by 10 percent, as
with our findings. Meghir and Palmer (1999) examined the impact of the Swedish school
reforms, i.e. the extension of compulsory schooling by one year, and this also corresponded
with our findings. Their result, obtained using the exogenous variation induced by reform
assignment, led to a point estimate that was higher than that derived using OLS, even when
they allowed for the heterogeneous returns to years of schooling. This is also consistent with
the idea that reform changed the composition of those taking higher education towards lower
average ability and poor family background. Ashenfelter et al. (1999) analysed several studies
in the US and seven non-US countries between 1974 and 1995. They found that IV and twin
study estimates exceeded OLS estimates by 3.1 and 1.6 percentage points respectively. But
after they controlled for studies that produced no interesting results and the insignificant
difference between the IV and the least-squares estimates, the differences were only 1.8 and
0.9 percentage points respectively (Fuente & Ciccone, 2002). Duflo (2001) examined the
effect of the school construction program in Indonesia on education and earnings. She found
returns to education ranging from 6.8 to 10.6 percent. Patrinos and Sakellariou (2004)
estimates for Venezuela found the private rate of return was 12 percent higher when using
compulsory education. Uusitalo (1999) used changes in education sector L/70 but the result
was not as significant as those for the UK or other countries. A few studies in urban China
also indicate that IV is higher than OLS by between 4 and 5 percentage points such as Giles et
al. (2004), Heckman & Li (2004) and Fleisher et al. (2005). However, they used family
background, quality of elementary education and other instruments related to socio-economic
-
21
indicators as the instrument. In contrast, some studies provide the opposite findings. For
example, Vieira (1999) considered legal changes in compulsory education in Portugal using
data drawn from Quadros de Pessoal for the years 1986 and 1992. The results showed high
standard errors and OLS estimates that were higher than those derived using IV.
Unfortunately, the comparison between OLS and IV estimations using Malaysian data with
different IV instruments could not be made because there have been no previous studies of
this kind relating to Malaysia. The main reason for this is probably the difficulty in data
availability.
7. CONCLUSION
Our estimate of the average private rate of return for an additional year of schooling in
Malaysia was 10.51 and 10.04 percent for 2002 and 2004, respectively. An additional year of
experience has increased earnings by 3 to 5 percent for all years of surveys. The Mincerian
human capital model, fitted well with the Malaysian data. The models coefficients and signs
were in line with the theory. The schooling parameters show the private rate of return to
education was similar to the world average and slightly higher than the average of Asia. The
estimation of the private rate of return to education using the IV approach is higher than
results from using OLS by approximately 10 to 11 percent. However, this result should be
interpreted carefully. Because either the impact of policy reforms or potential bias of OLS
estimation could effected the results of estimation. If the OLS estimation is consistent,
therefore, the higher returns estimated by IV reflect the impact of the school reform of 1970
on returns to education. On the other hand, if one considered the hypothesis that OLS
underestimated the return, then, the school reform of 1970 is a good instrument for the IV
-
22
approach. Essentially, it is impossible to separate between the methodological and exogenous
impact in estimating the returns to education.
To sum up, the findings of this study are as follows. Firstly, I found the average
private rate of returns for Malaysia to be almost consistent with the world average. The
homogenous return was about 10 percent, equal to the world average and slightly higher than
the Asian average. Secondly, the most important finding is that the returns to schooling in
Malaysia are best characterised by the heterogeneous returns model, implying that returns
vary across individuals. By using the IV method, I estimated the LATE from schooling
reforms and found the returns to be higher than the estimate using OLS. Using two sets of
data, HIS2002 and HIS2004, the private rate of return to education increased between 10 and
15 percent compared to conventional OLS-based estimation. It is likely that the difference in
estimates of rates of return to education using the OLS and the IV methods are not solely due
to the well-known tendency for IV to result in higher estimates than OLS. It is also likely due
to the fact that the schooling reform that is the chosen instrument in the IV method has itself
generated higher positive returns.
-
References
Angrist, J. D. & Krueger, A., 1992, The Effect of age at School Entry on Educational
Attainment: An Application of Instrumental Variables with Moments From two
samples. Journal of the American Statistical Association, 87, 328-336.
Ashenfelter, O., Harmon, C. & Oosterbeek, H., 1999, A review of estimates of the
schooling/earnings relationship, with test for publication bias. Labour Economics, 6,
(1999), 453-470.
Blau, D. M., 1986, Self-employment, Earnings, and Mobility in Peninsular Malaysia. World
Development, Vol. 14, No. 7, pp. 839-852.
Blundell, R., Dearden, L. and Sianesi, B., 2004, Evaluating the Impact of Education on
Earnings in the UK: Models, Methods and Results from the NCDS. University
College London and Institute of Fiscal Studies, December 2004, London.
Blundell, R., Dearden, L., Goodman, A and Reed, H., 2000, The Returns to Higher Education
in Britain: Evidence From a British Cohort. The Economic Journal, 110 (February),
F82 F99, Royal Economic Society.
Bound, J., Jeagar, D.A. and Baker, R. 1995, Problem with instrumental variables estimation
when the correlation between instruments and endogenous explanatory variables is
weak. Journal of the American Statistical Association, 90 (430), pp. 443-450.
Brunello, G. and Miniaci, R., 1999, The economics returns to schooling for Italian men. An
evaluation based in instrumentals variables. Labour Economics, 6, (1999), 509-519.
Card, D. 1999, The Causal Effect on Education and Earning, in O Aschenfelter & D Card
(ed.), Handbooks of Labour Economics 3, North-Holland.
Card, D. 2001, Estimating the Returns to Schooling: Progress on Some Persistent
Econometrics Problem. Econometrica, 69 (5), 1127-1160.
Chapman, B. J. and Harding, J. R., 1985, Sex Differing in Earnings: An Analysis of Malaysia
Wage Data. The Journal of Development Studies, 21, No. 3 (April 1985), 3672-73.
Chung, T. P. 2003, Returns to education: updates for Malaysia. Applied Economics Letters,
10, 837-841.
Chung, T. P. 2004, The Returns to Education and Training: Evidence from the Malaysian
Family Life Surveys. Pacific Economic Review, 9: 2 (2004), pp. 103-116.
Duflo, E. 2001, Schooling and Labor Market Consequences of School Construction in
Indonesia: Evidence from an Unusual Policy Experiment. American Economic
Review 91, no. 4: 795813.
Fleisher, B. M., Sirianova, K. and Wang, X., 2005, Returns to schooling in China under
planning and reform. Journal of Comparative Economics 33 (2005) 265-277.
-
24
Fuente, A. and Ciccone, A., 2002, Human capital in a global and knowledge-based economy.
OECD: Final Report, Paris.
Gallup, J. L. 1997, Ethnicity and earnings in Malaysia. Harvard Institute for International
Development, Development Discussion Paper No.592.
Giles, J., Park, A. and Zang, J. 2004, The proletarian cultural revolution, disruptions to
education, and returns to schooling in urban China. Department of Economic,
University of Michigan, An Arbor, MI.
Griliches, Z., 1977, Estimating the Returns to Schooling: Some Econometric Problems.
Econometrica, Vol. 45, No. 1 (Jan., 1977), 1-22.
Harmon, C. and Walker, I., 1995, Estimates of the Economic Return to Schooling for United
Kingdom. The American Economic Review, Vol.85, No. 5 (Dec., 1995), 1278-1286.
Hawley, J. D. 2004, Changing returns to education in times of prosperity and crisis, Thailand
19851998. Economics of Education Review, 23, 273-286.
Heckman, J. J., and Li, X., 2004, Selection bias, comparative advantages, heterogeneous
returns to schooling to education: Evidence from China in 2000. Pacific Economic
Review 9, 155-171.
Heckman, J.J. and Vytlacil, E.J., 2000, The relationship between treatment parameters within
a latent variable framework. Economics Letters 66 (2000) 3339.
Hoerr, O. D., 1973, Education, Income, and Equity in Malaysia. Economic Development and
Cultural Change, vol. 21, no. 2 (Jan., 1973), pp. 247-273.
Imbens, G. W. and Angrist, J. D., 1994, Identification and Estimation of Local Average
Treatment Effects. Econometrica, Vol. 62, No. 2. (Mar. 1994), pp. 467-475.
Lee, K. H., 1980, Education, earnings, and occupational status in Malaysia, 1978, PhD
thesis, Department of Economics, London School of Economics and Political
Science, University of London.
Mazumdar, D., 1981, The Urban Labor Market and Income Distribution: A study of
Malaysia. Oxford University Press, New York.
Meghir, C. and Palme, M., 1999, Assessing the Effect of Schooling on Earnings Using a
Social Experiment. Working Paper Series in Economics and Finance 313, Stockholm
School of Economics.
Mincer, J., 1974, Schooling, Experience and Earning. Columbia University Press, New York.
Ministry of Education, Malaysia 2001, Education in Malaysia. Dewan Bahasa dan Pustaka,
Kuala Lumpur.
Ministry of Education, Malaysia 1980, Education in Malaysia. Dewan Bahasa dan Pustaka,
Kuala Lumpur.
-
25
Moock, P. R., Patrinos, H. A. and Venkataraman, M. 2003, Education and earnings in a
transition economy: the case of Vietnam. Journal of Education Review, 22 (2003)
503-510.
Neville, W., 1998, Restructuring Education in Malaysia: the nature and applications of policy
changes. Higher Education Policy II (1998), 257- 279.
Okposin, S. B., Abdul Hamid, A. H. and Boon, H. O., 2005, The Changing Phase of
Malaysian Economy. Pelanduk Publications, Kuala Lumpur.
Patrinos, H. A. and Sakellariou, C., 2004, Schooling and Labor Market Impacts of a Natural
Policy Experiment. Policy Research Working Paper 3459, World Bank.
Pons, E. and Gonzalo, M.T., 2001, Returns to schooling in Spain. How reliable are IV
estimates?., Working Paper No. 446, November 2001, Queen Mary, University of
London.
Psacharopoulos, G. and Patrinos, H. A., 2002, Returns to Investment in Education: A Further
Update. World Bank Policy Research Working Paper 2881, September 2002, Latin
America and the Caribbean Region, World Bank.
Psacharopoulos, G. and Patrinos, H. A., 2004, Returns to Investment in Education: A Further
Update. Education Economics, Vol. 12, No. 2, August 2004.
Psacharopoulos, G., 1994, Returns to Investment in Education: A Global Update. World
Development 22(9):1325-43.
Rahmah, I. and Rogayah, M. Z., 2003, Earnings differentials determinants between skills in
the Malaysian manufacturing sector. Asian Economic Journal 2003, Vol. 17 No. 4,
325-340.
Ruppert, E., 1998, Managing foreign labour in Singapore and Malaysia: are there lessons for
GCC countries? World Bank Working Paper No.2053, World Bank.
Ryoo, J., Nam, Y-S. and Carnoy, M., 1993, Changing Rates of Return to Education over
Time: A Korean Case Study. Economics of Education Review, Vol. 12(1), pp.71-80.
Sakellariou, C., 2003, Returns to Formal and Technical Education in Singapore. Education
Economics, 11(1): 73-87.
Schultz, T. P. 1988, Education Investment and Returns, in H Chenery and TN Srinivasan
(ed.). Handbook of Development Economics, Volume I, Elsevier Science Publisher
B.V., Amsterdam.
Uusitalo, R., 1999, Return to education in Finlan. Labour Economics, 6 (1999), 569-580.
Vieira, J., 1999, Return to Education in Portugal. Labour Economics 6 (1999), 535-541.
Willis, R. J., 1986, Wages determinants, in O Eshenfelter & R Layard (ed.). Handbooks of
Labor Economics, vol. 1, Elsevier Science Publishers B.V, North Holland.